It seems that current Wikisource scan image handling implementation is terribly resource consumeng (network, memory):

While earlier scan image loaded were max 1k px width, now there are two images downloaded for each scan: max-width + 2*max-width (scalled up for some reason). This means, eg. assuming 2k-width scan (which is quite standard, now) that you will need to download and handle about 4^2 + 2^2 = 20 times larger network trafic (100% max-width + 200% max-width images).

This might SIGNIFICANTLY affect users who pay for network transfer.

And I wonder how was this accepted without wide discussion while earlier implementing prev/next image preload/prefetch discussion met an oposition from some developers who suggested that 10-100% (1.1-2x) larger network trafic could be a problem for some users.

Unfortunately, I am out of time resources today, so I would welcome if anybody fills a bug (or bugs) concerning this.

Ankry

PS. Setting scan width in index pages (likely to half of the required width) would probably be a temporary workaround for affected users.

I think most Wikisource developers are likely to be on this list. Of course, it's best to make sure there are Phabricator tickets for every separate bug or feature request.

On 21/11/21 1:36 am, Ankry wrote:

Well, I was notified by techncally skilled users that the ned OpenSeadragon library is much heavier and more memory consuming than curreently used tools. So I can only hope that its load into memory can be disabled if one needs so.

(may be critical while working on multiple pages at once)

However, I doubt if any technical comments from communities expressed here will reach developers. And which wiki pages would be more appropriate for such comments.

Ankry

W dniu 20.11.2021 o 14:33, Ruthven pisze:

Hi all,

  as usual, I get surprised every time there are major changes on the MediaWiki software that are deployed without providing advance warning to the community.

Every time it's the same story: something stops working on the project. A gadget, a toolbar or some personalised JS.

This time it was T288141 (see https://phabricator.wikimedia.org/T288141), that was deployed in all the Wikisources (then rolled back because WikiMedia computer scientists are the best) completely disrupting redesigning the image side of the Page namespace. This affected the toolbars (see https://phabricator.wikimedia.org/T296033) and several gadgets around all the Wikisources.

I am not saying that MediaWiki software shouldn't be improved: it's normal that we're trying to get all we can from this outdated software. I am just asking that major changes that affect all the Wikisources should be announced in every single Village Pump waaay before deploying them on the projects.

Is it possible, as a Usergroup, to do a little pressure to be considered as a community and not as guinea pigs on which to deploy new, partially-tested features?

Alex

Ruthven on Wikipedia

_______________________________________________
Wikisource-l mailing list -- wikisource-l@lists.wikimedia.org
To unsubscribe send an email to wikisource-l-leave@lists.wikimedia.org

_______________________________________________
Wikisource-l mailing list -- wikisource-l@lists.wikimedia.org
To unsubscribe send an email to wikisource-l-leave@lists.wikimedia.org

_______________________________________________
Wikisource-l mailing list -- wikisource-l@lists.wikimedia.org
To unsubscribe send an email to wikisource-l-leave@lists.wikimedia.org